View
2
Download
0
Category
Preview:
Citation preview
RIPE73
Measuring the IPv4 transfer markets
1
(Simula Research Laboratory) (CAIDA/UCSD)Ioana Livadariu, Ahmed Elmokashfi Amogh Dhamdhere
RIPE73
IPv4 address space: status
Available IPv4 /8 address blocks per RIR*
* Source: NRO Internet Number Resource Report June 2016
• Limited address space assignment • Organizations have to justify the required IPv4 address space
2
Num
ber
of /
8 ad
dres
s bl
ocks
RIPE73 3
IPv4 address space: allocation
IPv4 Address space management
RIPE73
IPv4 Transfers
• IPv4 address transactions that occur between organizations
• Can involve third-parties (IPv4 brokers) • Regulated by the RIRs policies
• Three RIRs have legitimized transfer markets
Internet Registry Intra-RIR Policy First Published Transfers
RIPE December 2008 October 2012
ARIN June 2009 October 2009
APNIC February 2010 January 2011
Internet Registry Inter-RIR Policy First Published Transfers ARIN <—> APNIC July 2012 October 2012
APNIC <—> RIPE <—> ARIN September 2015 December 2015
4
RIPE73
Outline
• Analysis of published transfers:
• How are transfers evolving over time ? • What type of addresses are being transferred ? • Are buyers using the acquired space ? • Is there a correlation between markets and IPv6
adoption ? • What is the market value ?
5
• Detecting transfers:
• Can we detect transfers using publicly available data ?
RIPE73
Increasing number of transfers over time
6
• 80% year on year increase (2013-2015); not much change in 2016 • Approx. 65% of the reported transfers occur within RIPE
0
100
200
300
400
500
600
700
800
900
Sep2009
Sep2010
Mar2011
Sep2011
Mar2012
Sep2012
Mar2013
Sep2013
Mar2014
Sep2014
Mar2015
Sep2015
Mar2016
Sep2016
Tran
sfer
s
Date
ARINAPNIC
RIPEInter-RIR
RIPE73
How much space is transferred?
7
• Transferred address blocks account for ~2.67% of the IPv4 space • 47% of the transferred space comes from ARIN
0x100
5x103
10x103
15x103
20x103
25x103
30x103
35x103
40x103
45x103
Sep2009
Sep2010
Mar2011
Sep2011
Mar2012
Sep2012
Mar2013
Sep2013
Mar2014
Sep2014
Mar2015
Sep2015
Mar2016
Sep2016
IPv4
add
ress
es (/
24s
Bloc
ks)
Date
ARINAPNIC
RIPEInter-RIR
RIPE73
Which space is being transferred?
8
• 71% of the transferred space is legacy allocation (>90% ARIN) • In ARIN, 37% of the blocks are larger than /20, whereas more than
80% transferred blocks are smaller than /20 for RIPE and APNIC
100
101
102
/10 /11 /12 /13 /14 /15 /16 /17 /18 /19 /20 /21 /22 /23 /24
Perc
enta
ge o
f tra
nsfe
rs
Block Size
ARINAPNIC
RIPEInter-RIR
RIPE73
Transferred space = “Used” space ?
Class Before After ererA Unrouted Unrouted
B Routed UnroutedC Unrouted RoutedD Routed Routed
94% of the transferred space is routed after the transfer
% (Total space)4.041.4985.179.27
Visibility of the transferred blocks in the routing table
RIR Time before re-announced (months) RIPE 1.91
APNIC 6.2ARIN 6.48
9
Buyers acquire addresses to meet immediate needs
RIPE73
Buyers need addresses more than sellers
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
T - 12 T - 9 T - 6 T - 3 T T + 3 T + 6 T + 9 T + 12
Util
izat
ion
fract
ion
(med
ian
valu
e)
T - x ( months )
ARINAPNIC
RIPEARIN-APNIC
* Source: ISI Census data
• Utilization fraction* = fraction of IP addresses that responds to ICMP requests in a transferred prefix
10
• Utilization fraction of the transferred space has increased with at least 50% after the transfer date
RIPE73 11
0
10
20
30
40
50
60
70
80
90
100
Mar2011
Sep2011
Mar2012
Sep2012
Mar2013
Sep2013
Mar2014
Sep2014
Mar2015
Sep2015
IPv4
add
ress
es (%
of /
24 B
lock
s pe
r RIR
)
Date
ARIN APNIC RIPE
High percentage of the address space is exchanged among the top participants in the market
Percentage of IPv4 address space sold/bought by top 10% dominant players per RIR
RIPE73
RIPE: Four countries dominate the market
(IPv4 space bought/IPv4 space sold) and IPv6 adoption* per country within RIPE
12
• Organizations involved in the IPv4 transfer market come from 64 countries
• Approx. 78% of the address space is exchanged between six countries
• 50% of the sold IPv4 space comes from two countries (Germany and Romania)
• 30% of the IPv4 space is bought by organizations in two countries (Saudi Arabia and Iran)
* Source: Google IPv6 Statistics
RIPE73
Are markets slowing down IPv6 adoption?
• Fraction of IPv6 adopters = fraction of buyers that are originating IPv6 prefixes* after acquiring IPv4 addresses on the transfer market
• Increasing number of buyers that adopt IPv6 across all RIRs • IPv4 transfers markets do not appear to inhibit the IPv6 adoption
13 * Source:BGP data (Routeviews and RIPE NCC)
0
0.1
0.2
0.3
0.4
0.5
0.6
Jan2009
Jul2009
Jan2010
Jul2010
Jan2011
Jul2011
Jan2012
Jul2012
Jan2013
Jul2013
Jan2014
Jul2014
Jan2015
Jul2015
Frac
tion
of IP
v6 a
dopt
ers
ARINAPNIC
RIPE
RIPE73
What is the market monetary value?
• Widely known IP transactions: 2011 Microsoft - Nortel*: $11 per IPv4 address 2011 Cerner - Borders** : $12 per IPv4 address
* Source: http://www.networkworld.com/
**Source: http://www.ipaddressnews.com/
IPv4 Market Group (retrieved in August 2015)
RIR /20 /19 /18 /17 /16APNIC 12.50 10.00 8.00 7.25 6.50ARIN 12.50 10.00 8.75 6.40 5.50RIPE 12.55 12.00 10.50 8.50 8.20
IPv4Auctions.com - (retrieved in September 2016)
14
• Prices published by IPv4 Brokers:
• Monetary aspects of the IPv4 transactions are confidential • Prices published by IPv4 Brokers offer a partial view of the market value
Block Sold Date Price per address/22 Block Registered in ARIN 9/20/16 $12.00
/21 Block Registered in ARIN 9/20/16 $10.00/23 Block Registered in ARIN 9/19/16 $12.00/24 Block Registered in ARIN 9/19/16 $13.35/23 Block Registered in ARIN 9/15/16 $11.75/21 Block Registered in ARIN 9/13/16 $11.15/23 Block Registered in ARIN 9/13/16 $12.74
/24 Block Registered in ARIN 9/13/16 $13.25/19 Block Registered in ARIN 8/24/16 $8.75
RIPE73
Estimating the market monetary value
• Approach: • We model the IPv4 address block prices using a Hedonic Pricing
Method*, which estimates the value of a good by taking into account both internal characteristics and external factors of the good
• We use prices reported by IPv4 brokers to fit the hedonic prices model
*Source: S. Rosen, Hedonic prices and implicit markets: Product differentiation in pure competition
• Estimated value of the market: ~$386 M (USD)
15
RIPE73
Can we detect transfers “in the wild” ?
• Transfers need to be approved by the RIRs, but there is no mechanism to ensure that organizations report to the RIRs
• Methodology: Use BGP data (routing table dumps) collected from January 2004 to September 2015 to construct prefix-AS mapping and identify prefixes that change origin AS
16
RIPE73
Detecting transfers: filtering approach
Filter Removed prefixes
Map2Organization IPv4 space movement within the same organization
Transient Short-lived advertised IPv4 space (e.g. prefix hijacks)
RIR IPv4 space advertised by RIRs
Delegation Provider-aggregatable address space
17
• Our approach is prone to false positives due to legitimate reasons • Design four filters to reduce the number of candidate transfers
RIPE73 18
Evaluating the results
• Filters reduce 65% of the initial number of candidate transfers • Our methodology infers more than 90% of the detectable
published transfers
102
103
104
Jan2004
Jan2005
Jan2006
Jan2007
Jan2008
Jan2009
Jan2010
Jan2011
Jan2012
Jan2013
Jan2014
Jan2015
Transfers
Date
InitialFinal
RIPE73 19
False positive: examples and causes
Non-BGP speaker organizations switching providers
Incomplete AS2ORG mapping*
Reallocated address blocks
*Source: CAIDA AS2ORG Mapping
RIPE73 20
Detecting transfers: expanding the methodology
• Possible solution: Augmenting the data (DNS names) • Changes in DNS resource record • Data: IPv4 Routed /24 DNS Names Dataset
• Preliminary analysis: Usage of DNS records removes two third of the analyzable candidate transfers
100
101
102
Jan2004
Jan2005
Jan2006
Jan2007
Jan2008
Jan2009
Jan2010
Jan2011
Jan2012
Jan2013
Jan2014
Jan2015
Tran
sfer
s (%
)
Date
Inferred (BGP-methodology)Final (ISC DNS data)
Final (ARK DNS data)
RIPE73
Conclusions
21
• Increase in the size of the IPv4 transfer markets • The majority of the transferred blocks are legacy
allocations • Markets seem to serve their intended purpose (i.e
buyers “use” the acquired address space) • Markets appear not to slow down the IPv6 adoption • Markets differ across regions in terms of size and type
of the transferred blocks, participants
• Detecting transfers is difficult and requires using multiple data sources
Recommended